Search academic texts

Information × Registration Number 0221U000021, 0120U103195 , R & D reports Title Biologically active compounds against tuberculosis: molecular basis, targeted search, biological screening. popup.stage_title Head Tukalo Michael A., Доктор біологічних наук Registration Date 02-01-2021 Organization Institute of Molecular Biology and Genetics of NAS of Ukraine popup.description2  As a result of the work, 24 models of neural networks (two- and four-class) and SVM models (models of the support-vector machine) were built: 8 two-class neural networks, 8 four-class neural networks and 8 SVM models. Their training and validation were performed on the basis of Mycobacterium tuberculosis methionyl-tRNA synthetase inhibitors. For SVM models, the ROC-score ranged from 0.95 to 0.97, and the MCC ranged from 0.71-0.84. For two-class neural networks, the ROC-score ranged from 0.87 to 0.95, and the MCC ranged from 0.61 to 0.86. For four-class neural networks, the ROC-score ranged from 0.79 to 0.84, and the MCC ranged from 0.34 to 0.43. Virtual screening of a compounds collection with more than 200,000 molecules was performed against the best models. According to the results of screening, 257 compounds were selected for biological testing for M. tuberculosis MetRS. Recombinant MetRSs of M. tuberculosis and human were obtained for biological tests. The constructed SVM models and neural networks can be used by other research groups in Ukraine and around the world for virtual screening to search for new classes of M. tuberculosis MetRS inhibitors. The obtained recombinant MetRSs of M. tuberculosis and humans will be used to test compounds with potential inhibitory activity against MetPCs selected by machine learning methods. Product Description popup.authors Bdzhola Volodymyr H. Hudzera Olha Yo. Kovalenko Oksana P. Protopopov Mykola V. Rybak Mariia Yu. Starosyla Serhii A. popup.nrat_date 2021-01-02 Close